The Dyad Ratios Algorithm for Estimating Latent Public Opinion

Abstract

In the study of longitudinal movements in public opinion it is usually the case that data are abundant, but irregular. Both cases (times) and variables are numerous, but it is never the case that all cases are available for any one variable. The dyad ratios algorithm was created to make dimensional analysis possible for this perverse data structure. The central logic of the dyad ratios algorithm is explained. Central focus is on the use of ratios as a starting point, the recursive estimation procedure, validity estimation by iterative procedure and the bootstrapping of standard errors. The ability of the algorithm to estimate a known longitudinal path is tested with artificial data. Then dyad ratios is compared to principal components analysis for a particular real data problem where both are possible. A final section makes a limited comparison between dyad ratios and item response theory.